linux_misc.c revision 293503
1/*- 2 * Copyright (c) 2002 Doug Rabson 3 * Copyright (c) 1994-1995 S��ren Schmidt 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer 11 * in this position and unchanged. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. The name of the author may not be used to endorse or promote products 16 * derived from this software without specific prior written permission 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 19 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 20 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 21 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 22 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 23 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 24 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 25 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 26 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 27 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 28 */ 29 30#include <sys/cdefs.h> 31__FBSDID("$FreeBSD: stable/10/sys/compat/linux/linux_misc.c 293503 2016-01-09 15:28:05Z dchagin $"); 32 33#include "opt_compat.h" 34#include "opt_kdtrace.h" 35 36#include <sys/param.h> 37#include <sys/blist.h> 38#include <sys/fcntl.h> 39#if defined(__i386__) 40#include <sys/imgact_aout.h> 41#endif 42#include <sys/jail.h> 43#include <sys/kernel.h> 44#include <sys/limits.h> 45#include <sys/lock.h> 46#include <sys/malloc.h> 47#include <sys/mman.h> 48#include <sys/mount.h> 49#include <sys/mutex.h> 50#include <sys/namei.h> 51#include <sys/priv.h> 52#include <sys/proc.h> 53#include <sys/reboot.h> 54#include <sys/racct.h> 55#include <sys/resourcevar.h> 56#include <sys/sched.h> 57#include <sys/signalvar.h> 58#include <sys/stat.h> 59#include <sys/syscallsubr.h> 60#include <sys/sysctl.h> 61#include <sys/sysproto.h> 62#include <sys/systm.h> 63#include <sys/time.h> 64#include <sys/vmmeter.h> 65#include <sys/vnode.h> 66#include <sys/wait.h> 67#include <sys/cpuset.h> 68 69#include <security/mac/mac_framework.h> 70 71#include <vm/vm.h> 72#include <vm/pmap.h> 73#include <vm/vm_kern.h> 74#include <vm/vm_map.h> 75#include <vm/vm_extern.h> 76#include <vm/vm_object.h> 77#include <vm/swap_pager.h> 78 79#ifdef COMPAT_LINUX32 80#include <machine/../linux32/linux.h> 81#include <machine/../linux32/linux32_proto.h> 82#else 83#include <machine/../linux/linux.h> 84#include <machine/../linux/linux_proto.h> 85#endif 86 87#include <compat/linux/linux_file.h> 88#include <compat/linux/linux_mib.h> 89#include <compat/linux/linux_signal.h> 90#include <compat/linux/linux_util.h> 91#include <compat/linux/linux_sysproto.h> 92#include <compat/linux/linux_emul.h> 93#include <compat/linux/linux_misc.h> 94 95int stclohz; /* Statistics clock frequency */ 96 97static unsigned int linux_to_bsd_resource[LINUX_RLIM_NLIMITS] = { 98 RLIMIT_CPU, RLIMIT_FSIZE, RLIMIT_DATA, RLIMIT_STACK, 99 RLIMIT_CORE, RLIMIT_RSS, RLIMIT_NPROC, RLIMIT_NOFILE, 100 RLIMIT_MEMLOCK, RLIMIT_AS 101}; 102 103struct l_sysinfo { 104 l_long uptime; /* Seconds since boot */ 105 l_ulong loads[3]; /* 1, 5, and 15 minute load averages */ 106#define LINUX_SYSINFO_LOADS_SCALE 65536 107 l_ulong totalram; /* Total usable main memory size */ 108 l_ulong freeram; /* Available memory size */ 109 l_ulong sharedram; /* Amount of shared memory */ 110 l_ulong bufferram; /* Memory used by buffers */ 111 l_ulong totalswap; /* Total swap space size */ 112 l_ulong freeswap; /* swap space still available */ 113 l_ushort procs; /* Number of current processes */ 114 l_ushort pads; 115 l_ulong totalbig; 116 l_ulong freebig; 117 l_uint mem_unit; 118 char _f[20-2*sizeof(l_long)-sizeof(l_int)]; /* padding */ 119}; 120int 121linux_sysinfo(struct thread *td, struct linux_sysinfo_args *args) 122{ 123 struct l_sysinfo sysinfo; 124 vm_object_t object; 125 int i, j; 126 struct timespec ts; 127 128 getnanouptime(&ts); 129 if (ts.tv_nsec != 0) 130 ts.tv_sec++; 131 sysinfo.uptime = ts.tv_sec; 132 133 /* Use the information from the mib to get our load averages */ 134 for (i = 0; i < 3; i++) 135 sysinfo.loads[i] = averunnable.ldavg[i] * 136 LINUX_SYSINFO_LOADS_SCALE / averunnable.fscale; 137 138 sysinfo.totalram = physmem * PAGE_SIZE; 139 sysinfo.freeram = sysinfo.totalram - cnt.v_wire_count * PAGE_SIZE; 140 141 sysinfo.sharedram = 0; 142 mtx_lock(&vm_object_list_mtx); 143 TAILQ_FOREACH(object, &vm_object_list, object_list) 144 if (object->shadow_count > 1) 145 sysinfo.sharedram += object->resident_page_count; 146 mtx_unlock(&vm_object_list_mtx); 147 148 sysinfo.sharedram *= PAGE_SIZE; 149 sysinfo.bufferram = 0; 150 151 swap_pager_status(&i, &j); 152 sysinfo.totalswap = i * PAGE_SIZE; 153 sysinfo.freeswap = (i - j) * PAGE_SIZE; 154 155 sysinfo.procs = nprocs; 156 157 /* The following are only present in newer Linux kernels. */ 158 sysinfo.totalbig = 0; 159 sysinfo.freebig = 0; 160 sysinfo.mem_unit = 1; 161 162 return (copyout(&sysinfo, args->info, sizeof(sysinfo))); 163} 164 165int 166linux_alarm(struct thread *td, struct linux_alarm_args *args) 167{ 168 struct itimerval it, old_it; 169 u_int secs; 170 int error; 171 172#ifdef DEBUG 173 if (ldebug(alarm)) 174 printf(ARGS(alarm, "%u"), args->secs); 175#endif 176 177 secs = args->secs; 178 179 if (secs > INT_MAX) 180 secs = INT_MAX; 181 182 it.it_value.tv_sec = (long) secs; 183 it.it_value.tv_usec = 0; 184 it.it_interval.tv_sec = 0; 185 it.it_interval.tv_usec = 0; 186 error = kern_setitimer(td, ITIMER_REAL, &it, &old_it); 187 if (error) 188 return (error); 189 if (timevalisset(&old_it.it_value)) { 190 if (old_it.it_value.tv_usec != 0) 191 old_it.it_value.tv_sec++; 192 td->td_retval[0] = old_it.it_value.tv_sec; 193 } 194 return (0); 195} 196 197int 198linux_brk(struct thread *td, struct linux_brk_args *args) 199{ 200 struct vmspace *vm = td->td_proc->p_vmspace; 201 vm_offset_t new, old; 202 struct obreak_args /* { 203 char * nsize; 204 } */ tmp; 205 206#ifdef DEBUG 207 if (ldebug(brk)) 208 printf(ARGS(brk, "%p"), (void *)(uintptr_t)args->dsend); 209#endif 210 old = (vm_offset_t)vm->vm_daddr + ctob(vm->vm_dsize); 211 new = (vm_offset_t)args->dsend; 212 tmp.nsize = (char *)new; 213 if (((caddr_t)new > vm->vm_daddr) && !sys_obreak(td, &tmp)) 214 td->td_retval[0] = (long)new; 215 else 216 td->td_retval[0] = (long)old; 217 218 return (0); 219} 220 221#if defined(__i386__) 222/* XXX: what about amd64/linux32? */ 223 224int 225linux_uselib(struct thread *td, struct linux_uselib_args *args) 226{ 227 struct nameidata ni; 228 struct vnode *vp; 229 struct exec *a_out; 230 struct vattr attr; 231 vm_offset_t vmaddr; 232 unsigned long file_offset; 233 unsigned long bss_size; 234 char *library; 235 ssize_t aresid; 236 int error, locked, writecount; 237 238 LCONVPATHEXIST(td, args->library, &library); 239 240#ifdef DEBUG 241 if (ldebug(uselib)) 242 printf(ARGS(uselib, "%s"), library); 243#endif 244 245 a_out = NULL; 246 locked = 0; 247 vp = NULL; 248 249 NDINIT(&ni, LOOKUP, ISOPEN | FOLLOW | LOCKLEAF | AUDITVNODE1, 250 UIO_SYSSPACE, library, td); 251 error = namei(&ni); 252 LFREEPATH(library); 253 if (error) 254 goto cleanup; 255 256 vp = ni.ni_vp; 257 NDFREE(&ni, NDF_ONLY_PNBUF); 258 259 /* 260 * From here on down, we have a locked vnode that must be unlocked. 261 * XXX: The code below largely duplicates exec_check_permissions(). 262 */ 263 locked = 1; 264 265 /* Writable? */ 266 error = VOP_GET_WRITECOUNT(vp, &writecount); 267 if (error != 0) 268 goto cleanup; 269 if (writecount != 0) { 270 error = ETXTBSY; 271 goto cleanup; 272 } 273 274 /* Executable? */ 275 error = VOP_GETATTR(vp, &attr, td->td_ucred); 276 if (error) 277 goto cleanup; 278 279 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) || 280 ((attr.va_mode & 0111) == 0) || (attr.va_type != VREG)) { 281 /* EACCESS is what exec(2) returns. */ 282 error = ENOEXEC; 283 goto cleanup; 284 } 285 286 /* Sensible size? */ 287 if (attr.va_size == 0) { 288 error = ENOEXEC; 289 goto cleanup; 290 } 291 292 /* Can we access it? */ 293 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td); 294 if (error) 295 goto cleanup; 296 297 /* 298 * XXX: This should use vn_open() so that it is properly authorized, 299 * and to reduce code redundancy all over the place here. 300 * XXX: Not really, it duplicates far more of exec_check_permissions() 301 * than vn_open(). 302 */ 303#ifdef MAC 304 error = mac_vnode_check_open(td->td_ucred, vp, VREAD); 305 if (error) 306 goto cleanup; 307#endif 308 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL); 309 if (error) 310 goto cleanup; 311 312 /* Pull in executable header into exec_map */ 313 error = vm_mmap(exec_map, (vm_offset_t *)&a_out, PAGE_SIZE, 314 VM_PROT_READ, VM_PROT_READ, 0, OBJT_VNODE, vp, 0); 315 if (error) 316 goto cleanup; 317 318 /* Is it a Linux binary ? */ 319 if (((a_out->a_magic >> 16) & 0xff) != 0x64) { 320 error = ENOEXEC; 321 goto cleanup; 322 } 323 324 /* 325 * While we are here, we should REALLY do some more checks 326 */ 327 328 /* Set file/virtual offset based on a.out variant. */ 329 switch ((int)(a_out->a_magic & 0xffff)) { 330 case 0413: /* ZMAGIC */ 331 file_offset = 1024; 332 break; 333 case 0314: /* QMAGIC */ 334 file_offset = 0; 335 break; 336 default: 337 error = ENOEXEC; 338 goto cleanup; 339 } 340 341 bss_size = round_page(a_out->a_bss); 342 343 /* Check various fields in header for validity/bounds. */ 344 if (a_out->a_text & PAGE_MASK || a_out->a_data & PAGE_MASK) { 345 error = ENOEXEC; 346 goto cleanup; 347 } 348 349 /* text + data can't exceed file size */ 350 if (a_out->a_data + a_out->a_text > attr.va_size) { 351 error = EFAULT; 352 goto cleanup; 353 } 354 355 /* 356 * text/data/bss must not exceed limits 357 * XXX - this is not complete. it should check current usage PLUS 358 * the resources needed by this library. 359 */ 360 PROC_LOCK(td->td_proc); 361 if (a_out->a_text > maxtsiz || 362 a_out->a_data + bss_size > lim_cur(td->td_proc, RLIMIT_DATA) || 363 racct_set(td->td_proc, RACCT_DATA, a_out->a_data + 364 bss_size) != 0) { 365 PROC_UNLOCK(td->td_proc); 366 error = ENOMEM; 367 goto cleanup; 368 } 369 PROC_UNLOCK(td->td_proc); 370 371 /* 372 * Prevent more writers. 373 * XXX: Note that if any of the VM operations fail below we don't 374 * clear this flag. 375 */ 376 VOP_SET_TEXT(vp); 377 378 /* 379 * Lock no longer needed 380 */ 381 locked = 0; 382 VOP_UNLOCK(vp, 0); 383 384 /* 385 * Check if file_offset page aligned. Currently we cannot handle 386 * misalinged file offsets, and so we read in the entire image 387 * (what a waste). 388 */ 389 if (file_offset & PAGE_MASK) { 390#ifdef DEBUG 391 printf("uselib: Non page aligned binary %lu\n", file_offset); 392#endif 393 /* Map text+data read/write/execute */ 394 395 /* a_entry is the load address and is page aligned */ 396 vmaddr = trunc_page(a_out->a_entry); 397 398 /* get anon user mapping, read+write+execute */ 399 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 400 &vmaddr, a_out->a_text + a_out->a_data, 0, VMFS_NO_SPACE, 401 VM_PROT_ALL, VM_PROT_ALL, 0); 402 if (error) 403 goto cleanup; 404 405 error = vn_rdwr(UIO_READ, vp, (void *)vmaddr, file_offset, 406 a_out->a_text + a_out->a_data, UIO_USERSPACE, 0, 407 td->td_ucred, NOCRED, &aresid, td); 408 if (error != 0) 409 goto cleanup; 410 if (aresid != 0) { 411 error = ENOEXEC; 412 goto cleanup; 413 } 414 } else { 415#ifdef DEBUG 416 printf("uselib: Page aligned binary %lu\n", file_offset); 417#endif 418 /* 419 * for QMAGIC, a_entry is 20 bytes beyond the load address 420 * to skip the executable header 421 */ 422 vmaddr = trunc_page(a_out->a_entry); 423 424 /* 425 * Map it all into the process's space as a single 426 * copy-on-write "data" segment. 427 */ 428 error = vm_mmap(&td->td_proc->p_vmspace->vm_map, &vmaddr, 429 a_out->a_text + a_out->a_data, VM_PROT_ALL, VM_PROT_ALL, 430 MAP_PRIVATE | MAP_FIXED, OBJT_VNODE, vp, file_offset); 431 if (error) 432 goto cleanup; 433 } 434#ifdef DEBUG 435 printf("mem=%08lx = %08lx %08lx\n", (long)vmaddr, ((long *)vmaddr)[0], 436 ((long *)vmaddr)[1]); 437#endif 438 if (bss_size != 0) { 439 /* Calculate BSS start address */ 440 vmaddr = trunc_page(a_out->a_entry) + a_out->a_text + 441 a_out->a_data; 442 443 /* allocate some 'anon' space */ 444 error = vm_map_find(&td->td_proc->p_vmspace->vm_map, NULL, 0, 445 &vmaddr, bss_size, 0, VMFS_NO_SPACE, VM_PROT_ALL, 446 VM_PROT_ALL, 0); 447 if (error) 448 goto cleanup; 449 } 450 451cleanup: 452 /* Unlock vnode if needed */ 453 if (locked) 454 VOP_UNLOCK(vp, 0); 455 456 /* Release the temporary mapping. */ 457 if (a_out) 458 kmap_free_wakeup(exec_map, (vm_offset_t)a_out, PAGE_SIZE); 459 460 return (error); 461} 462 463#endif /* __i386__ */ 464 465int 466linux_select(struct thread *td, struct linux_select_args *args) 467{ 468 l_timeval ltv; 469 struct timeval tv0, tv1, utv, *tvp; 470 int error; 471 472#ifdef DEBUG 473 if (ldebug(select)) 474 printf(ARGS(select, "%d, %p, %p, %p, %p"), args->nfds, 475 (void *)args->readfds, (void *)args->writefds, 476 (void *)args->exceptfds, (void *)args->timeout); 477#endif 478 479 /* 480 * Store current time for computation of the amount of 481 * time left. 482 */ 483 if (args->timeout) { 484 if ((error = copyin(args->timeout, <v, sizeof(ltv)))) 485 goto select_out; 486 utv.tv_sec = ltv.tv_sec; 487 utv.tv_usec = ltv.tv_usec; 488#ifdef DEBUG 489 if (ldebug(select)) 490 printf(LMSG("incoming timeout (%jd/%ld)"), 491 (intmax_t)utv.tv_sec, utv.tv_usec); 492#endif 493 494 if (itimerfix(&utv)) { 495 /* 496 * The timeval was invalid. Convert it to something 497 * valid that will act as it does under Linux. 498 */ 499 utv.tv_sec += utv.tv_usec / 1000000; 500 utv.tv_usec %= 1000000; 501 if (utv.tv_usec < 0) { 502 utv.tv_sec -= 1; 503 utv.tv_usec += 1000000; 504 } 505 if (utv.tv_sec < 0) 506 timevalclear(&utv); 507 } 508 microtime(&tv0); 509 tvp = &utv; 510 } else 511 tvp = NULL; 512 513 error = kern_select(td, args->nfds, args->readfds, args->writefds, 514 args->exceptfds, tvp, sizeof(l_int) * 8); 515 516#ifdef DEBUG 517 if (ldebug(select)) 518 printf(LMSG("real select returns %d"), error); 519#endif 520 if (error) 521 goto select_out; 522 523 if (args->timeout) { 524 if (td->td_retval[0]) { 525 /* 526 * Compute how much time was left of the timeout, 527 * by subtracting the current time and the time 528 * before we started the call, and subtracting 529 * that result from the user-supplied value. 530 */ 531 microtime(&tv1); 532 timevalsub(&tv1, &tv0); 533 timevalsub(&utv, &tv1); 534 if (utv.tv_sec < 0) 535 timevalclear(&utv); 536 } else 537 timevalclear(&utv); 538#ifdef DEBUG 539 if (ldebug(select)) 540 printf(LMSG("outgoing timeout (%jd/%ld)"), 541 (intmax_t)utv.tv_sec, utv.tv_usec); 542#endif 543 ltv.tv_sec = utv.tv_sec; 544 ltv.tv_usec = utv.tv_usec; 545 if ((error = copyout(<v, args->timeout, sizeof(ltv)))) 546 goto select_out; 547 } 548 549select_out: 550#ifdef DEBUG 551 if (ldebug(select)) 552 printf(LMSG("select_out -> %d"), error); 553#endif 554 return (error); 555} 556 557int 558linux_mremap(struct thread *td, struct linux_mremap_args *args) 559{ 560 struct munmap_args /* { 561 void *addr; 562 size_t len; 563 } */ bsd_args; 564 int error = 0; 565 566#ifdef DEBUG 567 if (ldebug(mremap)) 568 printf(ARGS(mremap, "%p, %08lx, %08lx, %08lx"), 569 (void *)(uintptr_t)args->addr, 570 (unsigned long)args->old_len, 571 (unsigned long)args->new_len, 572 (unsigned long)args->flags); 573#endif 574 575 if (args->flags & ~(LINUX_MREMAP_FIXED | LINUX_MREMAP_MAYMOVE)) { 576 td->td_retval[0] = 0; 577 return (EINVAL); 578 } 579 580 /* 581 * Check for the page alignment. 582 * Linux defines PAGE_MASK to be FreeBSD ~PAGE_MASK. 583 */ 584 if (args->addr & PAGE_MASK) { 585 td->td_retval[0] = 0; 586 return (EINVAL); 587 } 588 589 args->new_len = round_page(args->new_len); 590 args->old_len = round_page(args->old_len); 591 592 if (args->new_len > args->old_len) { 593 td->td_retval[0] = 0; 594 return (ENOMEM); 595 } 596 597 if (args->new_len < args->old_len) { 598 bsd_args.addr = 599 (caddr_t)((uintptr_t)args->addr + args->new_len); 600 bsd_args.len = args->old_len - args->new_len; 601 error = sys_munmap(td, &bsd_args); 602 } 603 604 td->td_retval[0] = error ? 0 : (uintptr_t)args->addr; 605 return (error); 606} 607 608#define LINUX_MS_ASYNC 0x0001 609#define LINUX_MS_INVALIDATE 0x0002 610#define LINUX_MS_SYNC 0x0004 611 612int 613linux_msync(struct thread *td, struct linux_msync_args *args) 614{ 615 struct msync_args bsd_args; 616 617 bsd_args.addr = (caddr_t)(uintptr_t)args->addr; 618 bsd_args.len = (uintptr_t)args->len; 619 bsd_args.flags = args->fl & ~LINUX_MS_SYNC; 620 621 return (sys_msync(td, &bsd_args)); 622} 623 624int 625linux_time(struct thread *td, struct linux_time_args *args) 626{ 627 struct timeval tv; 628 l_time_t tm; 629 int error; 630 631#ifdef DEBUG 632 if (ldebug(time)) 633 printf(ARGS(time, "*")); 634#endif 635 636 microtime(&tv); 637 tm = tv.tv_sec; 638 if (args->tm && (error = copyout(&tm, args->tm, sizeof(tm)))) 639 return (error); 640 td->td_retval[0] = tm; 641 return (0); 642} 643 644struct l_times_argv { 645 l_clock_t tms_utime; 646 l_clock_t tms_stime; 647 l_clock_t tms_cutime; 648 l_clock_t tms_cstime; 649}; 650 651 652/* 653 * Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value. 654 * Since 2.2.1 Glibc uses value exported from kernel via AT_CLKTCK 655 * auxiliary vector entry. 656 */ 657#define CLK_TCK 100 658 659#define CONVOTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 660#define CONVNTCK(r) (r.tv_sec * stclohz + r.tv_usec / (1000000 / stclohz)) 661 662#define CONVTCK(r) (linux_kernver(td) >= LINUX_KERNVER_2004000 ? \ 663 CONVNTCK(r) : CONVOTCK(r)) 664 665int 666linux_times(struct thread *td, struct linux_times_args *args) 667{ 668 struct timeval tv, utime, stime, cutime, cstime; 669 struct l_times_argv tms; 670 struct proc *p; 671 int error; 672 673#ifdef DEBUG 674 if (ldebug(times)) 675 printf(ARGS(times, "*")); 676#endif 677 678 if (args->buf != NULL) { 679 p = td->td_proc; 680 PROC_LOCK(p); 681 PROC_STATLOCK(p); 682 calcru(p, &utime, &stime); 683 PROC_STATUNLOCK(p); 684 calccru(p, &cutime, &cstime); 685 PROC_UNLOCK(p); 686 687 tms.tms_utime = CONVTCK(utime); 688 tms.tms_stime = CONVTCK(stime); 689 690 tms.tms_cutime = CONVTCK(cutime); 691 tms.tms_cstime = CONVTCK(cstime); 692 693 if ((error = copyout(&tms, args->buf, sizeof(tms)))) 694 return (error); 695 } 696 697 microuptime(&tv); 698 td->td_retval[0] = (int)CONVTCK(tv); 699 return (0); 700} 701 702int 703linux_newuname(struct thread *td, struct linux_newuname_args *args) 704{ 705 struct l_new_utsname utsname; 706 char osname[LINUX_MAX_UTSNAME]; 707 char osrelease[LINUX_MAX_UTSNAME]; 708 char *p; 709 710#ifdef DEBUG 711 if (ldebug(newuname)) 712 printf(ARGS(newuname, "*")); 713#endif 714 715 linux_get_osname(td, osname); 716 linux_get_osrelease(td, osrelease); 717 718 bzero(&utsname, sizeof(utsname)); 719 strlcpy(utsname.sysname, osname, LINUX_MAX_UTSNAME); 720 getcredhostname(td->td_ucred, utsname.nodename, LINUX_MAX_UTSNAME); 721 getcreddomainname(td->td_ucred, utsname.domainname, LINUX_MAX_UTSNAME); 722 strlcpy(utsname.release, osrelease, LINUX_MAX_UTSNAME); 723 strlcpy(utsname.version, version, LINUX_MAX_UTSNAME); 724 for (p = utsname.version; *p != '\0'; ++p) 725 if (*p == '\n') { 726 *p = '\0'; 727 break; 728 } 729 strlcpy(utsname.machine, linux_platform, LINUX_MAX_UTSNAME); 730 731 return (copyout(&utsname, args->buf, sizeof(utsname))); 732} 733 734#if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32)) 735struct l_utimbuf { 736 l_time_t l_actime; 737 l_time_t l_modtime; 738}; 739 740int 741linux_utime(struct thread *td, struct linux_utime_args *args) 742{ 743 struct timeval tv[2], *tvp; 744 struct l_utimbuf lut; 745 char *fname; 746 int error; 747 748 LCONVPATHEXIST(td, args->fname, &fname); 749 750#ifdef DEBUG 751 if (ldebug(utime)) 752 printf(ARGS(utime, "%s, *"), fname); 753#endif 754 755 if (args->times) { 756 if ((error = copyin(args->times, &lut, sizeof lut))) { 757 LFREEPATH(fname); 758 return (error); 759 } 760 tv[0].tv_sec = lut.l_actime; 761 tv[0].tv_usec = 0; 762 tv[1].tv_sec = lut.l_modtime; 763 tv[1].tv_usec = 0; 764 tvp = tv; 765 } else 766 tvp = NULL; 767 768 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 769 LFREEPATH(fname); 770 return (error); 771} 772 773int 774linux_utimes(struct thread *td, struct linux_utimes_args *args) 775{ 776 l_timeval ltv[2]; 777 struct timeval tv[2], *tvp = NULL; 778 char *fname; 779 int error; 780 781 LCONVPATHEXIST(td, args->fname, &fname); 782 783#ifdef DEBUG 784 if (ldebug(utimes)) 785 printf(ARGS(utimes, "%s, *"), fname); 786#endif 787 788 if (args->tptr != NULL) { 789 if ((error = copyin(args->tptr, ltv, sizeof ltv))) { 790 LFREEPATH(fname); 791 return (error); 792 } 793 tv[0].tv_sec = ltv[0].tv_sec; 794 tv[0].tv_usec = ltv[0].tv_usec; 795 tv[1].tv_sec = ltv[1].tv_sec; 796 tv[1].tv_usec = ltv[1].tv_usec; 797 tvp = tv; 798 } 799 800 error = kern_utimes(td, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 801 LFREEPATH(fname); 802 return (error); 803} 804 805int 806linux_futimesat(struct thread *td, struct linux_futimesat_args *args) 807{ 808 l_timeval ltv[2]; 809 struct timeval tv[2], *tvp = NULL; 810 char *fname; 811 int error, dfd; 812 813 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 814 LCONVPATHEXIST_AT(td, args->filename, &fname, dfd); 815 816#ifdef DEBUG 817 if (ldebug(futimesat)) 818 printf(ARGS(futimesat, "%s, *"), fname); 819#endif 820 821 if (args->utimes != NULL) { 822 if ((error = copyin(args->utimes, ltv, sizeof ltv))) { 823 LFREEPATH(fname); 824 return (error); 825 } 826 tv[0].tv_sec = ltv[0].tv_sec; 827 tv[0].tv_usec = ltv[0].tv_usec; 828 tv[1].tv_sec = ltv[1].tv_sec; 829 tv[1].tv_usec = ltv[1].tv_usec; 830 tvp = tv; 831 } 832 833 error = kern_utimesat(td, dfd, fname, UIO_SYSSPACE, tvp, UIO_SYSSPACE); 834 LFREEPATH(fname); 835 return (error); 836} 837#endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */ 838 839int 840linux_common_wait(struct thread *td, int pid, int *status, 841 int options, struct rusage *ru) 842{ 843 int error, tmpstat; 844 845 error = kern_wait(td, pid, &tmpstat, options, ru); 846 if (error) 847 return (error); 848 849 if (status) { 850 tmpstat &= 0xffff; 851 if (WIFSIGNALED(tmpstat)) 852 tmpstat = (tmpstat & 0xffffff80) | 853 BSD_TO_LINUX_SIGNAL(WTERMSIG(tmpstat)); 854 else if (WIFSTOPPED(tmpstat)) 855 tmpstat = (tmpstat & 0xffff00ff) | 856 (BSD_TO_LINUX_SIGNAL(WSTOPSIG(tmpstat)) << 8); 857 error = copyout(&tmpstat, status, sizeof(int)); 858 } 859 860 return (error); 861} 862 863int 864linux_waitpid(struct thread *td, struct linux_waitpid_args *args) 865{ 866 int options; 867 868#ifdef DEBUG 869 if (ldebug(waitpid)) 870 printf(ARGS(waitpid, "%d, %p, %d"), 871 args->pid, (void *)args->status, args->options); 872#endif 873 /* 874 * this is necessary because the test in kern_wait doesn't work 875 * because we mess with the options here 876 */ 877 if (args->options & ~(WUNTRACED | WNOHANG | WCONTINUED | __WCLONE)) 878 return (EINVAL); 879 880 options = (args->options & (WNOHANG | WUNTRACED)); 881 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ 882 if (args->options & __WCLONE) 883 options |= WLINUXCLONE; 884 885 return (linux_common_wait(td, args->pid, args->status, options, NULL)); 886} 887 888int 889linux_wait4(struct thread *td, struct linux_wait4_args *args) 890{ 891 int error, options; 892 struct rusage ru, *rup; 893 894#ifdef DEBUG 895 if (ldebug(wait4)) 896 printf(ARGS(wait4, "%d, %p, %d, %p"), 897 args->pid, (void *)args->status, args->options, 898 (void *)args->rusage); 899#endif 900 901 options = (args->options & (WNOHANG | WUNTRACED)); 902 /* WLINUXCLONE should be equal to __WCLONE, but we make sure */ 903 if (args->options & __WCLONE) 904 options |= WLINUXCLONE; 905 906 if (args->rusage != NULL) 907 rup = &ru; 908 else 909 rup = NULL; 910 error = linux_common_wait(td, args->pid, args->status, options, rup); 911 if (error != 0) 912 return (error); 913 if (args->rusage != NULL) 914 error = linux_copyout_rusage(&ru, args->rusage); 915 return (error); 916} 917 918int 919linux_waitid(struct thread *td, struct linux_waitid_args *args) 920{ 921 int status, options, sig; 922 struct __wrusage wru; 923 siginfo_t siginfo; 924 l_siginfo_t lsi; 925 idtype_t idtype; 926 struct proc *p; 927 int error; 928 929 options = 0; 930 linux_to_bsd_waitopts(args->options, &options); 931 932 if (options & ~(WNOHANG | WNOWAIT | WEXITED | WUNTRACED | WCONTINUED)) 933 return (EINVAL); 934 if (!(options & (WEXITED | WUNTRACED | WCONTINUED))) 935 return (EINVAL); 936 937 switch (args->idtype) { 938 case LINUX_P_ALL: 939 idtype = P_ALL; 940 break; 941 case LINUX_P_PID: 942 if (args->id <= 0) 943 return (EINVAL); 944 idtype = P_PID; 945 break; 946 case LINUX_P_PGID: 947 if (args->id <= 0) 948 return (EINVAL); 949 idtype = P_PGID; 950 break; 951 default: 952 return (EINVAL); 953 } 954 955 error = kern_wait6(td, idtype, args->id, &status, options, 956 &wru, &siginfo); 957 if (error != 0) 958 return (error); 959 if (args->rusage != NULL) { 960 error = linux_copyout_rusage(&wru.wru_children, 961 args->rusage); 962 if (error != 0) 963 return (error); 964 } 965 if (args->info != NULL) { 966 p = td->td_proc; 967 if (td->td_retval[0] == 0) 968 bzero(&lsi, sizeof(lsi)); 969 else { 970 sig = BSD_TO_LINUX_SIGNAL(siginfo.si_signo); 971 siginfo_to_lsiginfo(&siginfo, &lsi, sig); 972 } 973 error = copyout(&lsi, args->info, sizeof(lsi)); 974 } 975 td->td_retval[0] = 0; 976 977 return (error); 978} 979 980int 981linux_mknod(struct thread *td, struct linux_mknod_args *args) 982{ 983 char *path; 984 int error; 985 986 LCONVPATHCREAT(td, args->path, &path); 987 988#ifdef DEBUG 989 if (ldebug(mknod)) 990 printf(ARGS(mknod, "%s, %d, %d"), path, args->mode, args->dev); 991#endif 992 993 switch (args->mode & S_IFMT) { 994 case S_IFIFO: 995 case S_IFSOCK: 996 error = kern_mkfifo(td, path, UIO_SYSSPACE, args->mode); 997 break; 998 999 case S_IFCHR: 1000 case S_IFBLK: 1001 error = kern_mknod(td, path, UIO_SYSSPACE, args->mode, 1002 args->dev); 1003 break; 1004 1005 case S_IFDIR: 1006 error = EPERM; 1007 break; 1008 1009 case 0: 1010 args->mode |= S_IFREG; 1011 /* FALLTHROUGH */ 1012 case S_IFREG: 1013 error = kern_open(td, path, UIO_SYSSPACE, 1014 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1015 if (error == 0) 1016 kern_close(td, td->td_retval[0]); 1017 break; 1018 1019 default: 1020 error = EINVAL; 1021 break; 1022 } 1023 LFREEPATH(path); 1024 return (error); 1025} 1026 1027int 1028linux_mknodat(struct thread *td, struct linux_mknodat_args *args) 1029{ 1030 char *path; 1031 int error, dfd; 1032 1033 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd; 1034 LCONVPATHCREAT_AT(td, args->filename, &path, dfd); 1035 1036#ifdef DEBUG 1037 if (ldebug(mknodat)) 1038 printf(ARGS(mknodat, "%s, %d, %d"), path, args->mode, args->dev); 1039#endif 1040 1041 switch (args->mode & S_IFMT) { 1042 case S_IFIFO: 1043 case S_IFSOCK: 1044 error = kern_mkfifoat(td, dfd, path, UIO_SYSSPACE, args->mode); 1045 break; 1046 1047 case S_IFCHR: 1048 case S_IFBLK: 1049 error = kern_mknodat(td, dfd, path, UIO_SYSSPACE, args->mode, 1050 args->dev); 1051 break; 1052 1053 case S_IFDIR: 1054 error = EPERM; 1055 break; 1056 1057 case 0: 1058 args->mode |= S_IFREG; 1059 /* FALLTHROUGH */ 1060 case S_IFREG: 1061 error = kern_openat(td, dfd, path, UIO_SYSSPACE, 1062 O_WRONLY | O_CREAT | O_TRUNC, args->mode); 1063 if (error == 0) 1064 kern_close(td, td->td_retval[0]); 1065 break; 1066 1067 default: 1068 error = EINVAL; 1069 break; 1070 } 1071 LFREEPATH(path); 1072 return (error); 1073} 1074 1075/* 1076 * UGH! This is just about the dumbest idea I've ever heard!! 1077 */ 1078int 1079linux_personality(struct thread *td, struct linux_personality_args *args) 1080{ 1081#ifdef DEBUG 1082 if (ldebug(personality)) 1083 printf(ARGS(personality, "%lu"), (unsigned long)args->per); 1084#endif 1085 if (args->per != 0) 1086 return (EINVAL); 1087 1088 /* Yes Jim, it's still a Linux... */ 1089 td->td_retval[0] = 0; 1090 return (0); 1091} 1092 1093struct l_itimerval { 1094 l_timeval it_interval; 1095 l_timeval it_value; 1096}; 1097 1098#define B2L_ITIMERVAL(bip, lip) \ 1099 (bip)->it_interval.tv_sec = (lip)->it_interval.tv_sec; \ 1100 (bip)->it_interval.tv_usec = (lip)->it_interval.tv_usec; \ 1101 (bip)->it_value.tv_sec = (lip)->it_value.tv_sec; \ 1102 (bip)->it_value.tv_usec = (lip)->it_value.tv_usec; 1103 1104int 1105linux_setitimer(struct thread *td, struct linux_setitimer_args *uap) 1106{ 1107 int error; 1108 struct l_itimerval ls; 1109 struct itimerval aitv, oitv; 1110 1111#ifdef DEBUG 1112 if (ldebug(setitimer)) 1113 printf(ARGS(setitimer, "%p, %p"), 1114 (void *)uap->itv, (void *)uap->oitv); 1115#endif 1116 1117 if (uap->itv == NULL) { 1118 uap->itv = uap->oitv; 1119 return (linux_getitimer(td, (struct linux_getitimer_args *)uap)); 1120 } 1121 1122 error = copyin(uap->itv, &ls, sizeof(ls)); 1123 if (error != 0) 1124 return (error); 1125 B2L_ITIMERVAL(&aitv, &ls); 1126#ifdef DEBUG 1127 if (ldebug(setitimer)) { 1128 printf("setitimer: value: sec: %jd, usec: %ld\n", 1129 (intmax_t)aitv.it_value.tv_sec, aitv.it_value.tv_usec); 1130 printf("setitimer: interval: sec: %jd, usec: %ld\n", 1131 (intmax_t)aitv.it_interval.tv_sec, aitv.it_interval.tv_usec); 1132 } 1133#endif 1134 error = kern_setitimer(td, uap->which, &aitv, &oitv); 1135 if (error != 0 || uap->oitv == NULL) 1136 return (error); 1137 B2L_ITIMERVAL(&ls, &oitv); 1138 1139 return (copyout(&ls, uap->oitv, sizeof(ls))); 1140} 1141 1142int 1143linux_getitimer(struct thread *td, struct linux_getitimer_args *uap) 1144{ 1145 int error; 1146 struct l_itimerval ls; 1147 struct itimerval aitv; 1148 1149#ifdef DEBUG 1150 if (ldebug(getitimer)) 1151 printf(ARGS(getitimer, "%p"), (void *)uap->itv); 1152#endif 1153 error = kern_getitimer(td, uap->which, &aitv); 1154 if (error != 0) 1155 return (error); 1156 B2L_ITIMERVAL(&ls, &aitv); 1157 return (copyout(&ls, uap->itv, sizeof(ls))); 1158} 1159 1160int 1161linux_nice(struct thread *td, struct linux_nice_args *args) 1162{ 1163 struct setpriority_args bsd_args; 1164 1165 bsd_args.which = PRIO_PROCESS; 1166 bsd_args.who = 0; /* current process */ 1167 bsd_args.prio = args->inc; 1168 return (sys_setpriority(td, &bsd_args)); 1169} 1170 1171int 1172linux_setgroups(struct thread *td, struct linux_setgroups_args *args) 1173{ 1174 struct ucred *newcred, *oldcred; 1175 l_gid_t *linux_gidset; 1176 gid_t *bsd_gidset; 1177 int ngrp, error; 1178 struct proc *p; 1179 1180 ngrp = args->gidsetsize; 1181 if (ngrp < 0 || ngrp >= ngroups_max + 1) 1182 return (EINVAL); 1183 linux_gidset = malloc(ngrp * sizeof(*linux_gidset), M_TEMP, M_WAITOK); 1184 error = copyin(args->grouplist, linux_gidset, ngrp * sizeof(l_gid_t)); 1185 if (error) 1186 goto out; 1187 newcred = crget(); 1188 p = td->td_proc; 1189 PROC_LOCK(p); 1190 oldcred = crcopysafe(p, newcred); 1191 1192 /* 1193 * cr_groups[0] holds egid. Setting the whole set from 1194 * the supplied set will cause egid to be changed too. 1195 * Keep cr_groups[0] unchanged to prevent that. 1196 */ 1197 1198 if ((error = priv_check_cred(oldcred, PRIV_CRED_SETGROUPS, 0)) != 0) { 1199 PROC_UNLOCK(p); 1200 crfree(newcred); 1201 goto out; 1202 } 1203 1204 if (ngrp > 0) { 1205 newcred->cr_ngroups = ngrp + 1; 1206 1207 bsd_gidset = newcred->cr_groups; 1208 ngrp--; 1209 while (ngrp >= 0) { 1210 bsd_gidset[ngrp + 1] = linux_gidset[ngrp]; 1211 ngrp--; 1212 } 1213 } else 1214 newcred->cr_ngroups = 1; 1215 1216 setsugid(p); 1217 p->p_ucred = newcred; 1218 PROC_UNLOCK(p); 1219 crfree(oldcred); 1220 error = 0; 1221out: 1222 free(linux_gidset, M_TEMP); 1223 return (error); 1224} 1225 1226int 1227linux_getgroups(struct thread *td, struct linux_getgroups_args *args) 1228{ 1229 struct ucred *cred; 1230 l_gid_t *linux_gidset; 1231 gid_t *bsd_gidset; 1232 int bsd_gidsetsz, ngrp, error; 1233 1234 cred = td->td_ucred; 1235 bsd_gidset = cred->cr_groups; 1236 bsd_gidsetsz = cred->cr_ngroups - 1; 1237 1238 /* 1239 * cr_groups[0] holds egid. Returning the whole set 1240 * here will cause a duplicate. Exclude cr_groups[0] 1241 * to prevent that. 1242 */ 1243 1244 if ((ngrp = args->gidsetsize) == 0) { 1245 td->td_retval[0] = bsd_gidsetsz; 1246 return (0); 1247 } 1248 1249 if (ngrp < bsd_gidsetsz) 1250 return (EINVAL); 1251 1252 ngrp = 0; 1253 linux_gidset = malloc(bsd_gidsetsz * sizeof(*linux_gidset), 1254 M_TEMP, M_WAITOK); 1255 while (ngrp < bsd_gidsetsz) { 1256 linux_gidset[ngrp] = bsd_gidset[ngrp + 1]; 1257 ngrp++; 1258 } 1259 1260 error = copyout(linux_gidset, args->grouplist, ngrp * sizeof(l_gid_t)); 1261 free(linux_gidset, M_TEMP); 1262 if (error) 1263 return (error); 1264 1265 td->td_retval[0] = ngrp; 1266 return (0); 1267} 1268 1269int 1270linux_setrlimit(struct thread *td, struct linux_setrlimit_args *args) 1271{ 1272 struct rlimit bsd_rlim; 1273 struct l_rlimit rlim; 1274 u_int which; 1275 int error; 1276 1277#ifdef DEBUG 1278 if (ldebug(setrlimit)) 1279 printf(ARGS(setrlimit, "%d, %p"), 1280 args->resource, (void *)args->rlim); 1281#endif 1282 1283 if (args->resource >= LINUX_RLIM_NLIMITS) 1284 return (EINVAL); 1285 1286 which = linux_to_bsd_resource[args->resource]; 1287 if (which == -1) 1288 return (EINVAL); 1289 1290 error = copyin(args->rlim, &rlim, sizeof(rlim)); 1291 if (error) 1292 return (error); 1293 1294 bsd_rlim.rlim_cur = (rlim_t)rlim.rlim_cur; 1295 bsd_rlim.rlim_max = (rlim_t)rlim.rlim_max; 1296 return (kern_setrlimit(td, which, &bsd_rlim)); 1297} 1298 1299int 1300linux_old_getrlimit(struct thread *td, struct linux_old_getrlimit_args *args) 1301{ 1302 struct l_rlimit rlim; 1303 struct proc *p = td->td_proc; 1304 struct rlimit bsd_rlim; 1305 u_int which; 1306 1307#ifdef DEBUG 1308 if (ldebug(old_getrlimit)) 1309 printf(ARGS(old_getrlimit, "%d, %p"), 1310 args->resource, (void *)args->rlim); 1311#endif 1312 1313 if (args->resource >= LINUX_RLIM_NLIMITS) 1314 return (EINVAL); 1315 1316 which = linux_to_bsd_resource[args->resource]; 1317 if (which == -1) 1318 return (EINVAL); 1319 1320 PROC_LOCK(p); 1321 lim_rlimit(p, which, &bsd_rlim); 1322 PROC_UNLOCK(p); 1323 1324#ifdef COMPAT_LINUX32 1325 rlim.rlim_cur = (unsigned int)bsd_rlim.rlim_cur; 1326 if (rlim.rlim_cur == UINT_MAX) 1327 rlim.rlim_cur = INT_MAX; 1328 rlim.rlim_max = (unsigned int)bsd_rlim.rlim_max; 1329 if (rlim.rlim_max == UINT_MAX) 1330 rlim.rlim_max = INT_MAX; 1331#else 1332 rlim.rlim_cur = (unsigned long)bsd_rlim.rlim_cur; 1333 if (rlim.rlim_cur == ULONG_MAX) 1334 rlim.rlim_cur = LONG_MAX; 1335 rlim.rlim_max = (unsigned long)bsd_rlim.rlim_max; 1336 if (rlim.rlim_max == ULONG_MAX) 1337 rlim.rlim_max = LONG_MAX; 1338#endif 1339 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1340} 1341 1342int 1343linux_getrlimit(struct thread *td, struct linux_getrlimit_args *args) 1344{ 1345 struct l_rlimit rlim; 1346 struct proc *p = td->td_proc; 1347 struct rlimit bsd_rlim; 1348 u_int which; 1349 1350#ifdef DEBUG 1351 if (ldebug(getrlimit)) 1352 printf(ARGS(getrlimit, "%d, %p"), 1353 args->resource, (void *)args->rlim); 1354#endif 1355 1356 if (args->resource >= LINUX_RLIM_NLIMITS) 1357 return (EINVAL); 1358 1359 which = linux_to_bsd_resource[args->resource]; 1360 if (which == -1) 1361 return (EINVAL); 1362 1363 PROC_LOCK(p); 1364 lim_rlimit(p, which, &bsd_rlim); 1365 PROC_UNLOCK(p); 1366 1367 rlim.rlim_cur = (l_ulong)bsd_rlim.rlim_cur; 1368 rlim.rlim_max = (l_ulong)bsd_rlim.rlim_max; 1369 return (copyout(&rlim, args->rlim, sizeof(rlim))); 1370} 1371 1372int 1373linux_sched_setscheduler(struct thread *td, 1374 struct linux_sched_setscheduler_args *args) 1375{ 1376 struct sched_param sched_param; 1377 struct thread *tdt; 1378 int error, policy; 1379 1380#ifdef DEBUG 1381 if (ldebug(sched_setscheduler)) 1382 printf(ARGS(sched_setscheduler, "%d, %d, %p"), 1383 args->pid, args->policy, (const void *)args->param); 1384#endif 1385 1386 switch (args->policy) { 1387 case LINUX_SCHED_OTHER: 1388 policy = SCHED_OTHER; 1389 break; 1390 case LINUX_SCHED_FIFO: 1391 policy = SCHED_FIFO; 1392 break; 1393 case LINUX_SCHED_RR: 1394 policy = SCHED_RR; 1395 break; 1396 default: 1397 return (EINVAL); 1398 } 1399 1400 error = copyin(args->param, &sched_param, sizeof(sched_param)); 1401 if (error) 1402 return (error); 1403 1404 tdt = linux_tdfind(td, args->pid, -1); 1405 if (tdt == NULL) 1406 return (ESRCH); 1407 1408 error = kern_sched_setscheduler(td, tdt, policy, &sched_param); 1409 PROC_UNLOCK(tdt->td_proc); 1410 return (error); 1411} 1412 1413int 1414linux_sched_getscheduler(struct thread *td, 1415 struct linux_sched_getscheduler_args *args) 1416{ 1417 struct thread *tdt; 1418 int error, policy; 1419 1420#ifdef DEBUG 1421 if (ldebug(sched_getscheduler)) 1422 printf(ARGS(sched_getscheduler, "%d"), args->pid); 1423#endif 1424 1425 tdt = linux_tdfind(td, args->pid, -1); 1426 if (tdt == NULL) 1427 return (ESRCH); 1428 1429 error = kern_sched_getscheduler(td, tdt, &policy); 1430 PROC_UNLOCK(tdt->td_proc); 1431 1432 switch (policy) { 1433 case SCHED_OTHER: 1434 td->td_retval[0] = LINUX_SCHED_OTHER; 1435 break; 1436 case SCHED_FIFO: 1437 td->td_retval[0] = LINUX_SCHED_FIFO; 1438 break; 1439 case SCHED_RR: 1440 td->td_retval[0] = LINUX_SCHED_RR; 1441 break; 1442 } 1443 return (error); 1444} 1445 1446int 1447linux_sched_get_priority_max(struct thread *td, 1448 struct linux_sched_get_priority_max_args *args) 1449{ 1450 struct sched_get_priority_max_args bsd; 1451 1452#ifdef DEBUG 1453 if (ldebug(sched_get_priority_max)) 1454 printf(ARGS(sched_get_priority_max, "%d"), args->policy); 1455#endif 1456 1457 switch (args->policy) { 1458 case LINUX_SCHED_OTHER: 1459 bsd.policy = SCHED_OTHER; 1460 break; 1461 case LINUX_SCHED_FIFO: 1462 bsd.policy = SCHED_FIFO; 1463 break; 1464 case LINUX_SCHED_RR: 1465 bsd.policy = SCHED_RR; 1466 break; 1467 default: 1468 return (EINVAL); 1469 } 1470 return (sys_sched_get_priority_max(td, &bsd)); 1471} 1472 1473int 1474linux_sched_get_priority_min(struct thread *td, 1475 struct linux_sched_get_priority_min_args *args) 1476{ 1477 struct sched_get_priority_min_args bsd; 1478 1479#ifdef DEBUG 1480 if (ldebug(sched_get_priority_min)) 1481 printf(ARGS(sched_get_priority_min, "%d"), args->policy); 1482#endif 1483 1484 switch (args->policy) { 1485 case LINUX_SCHED_OTHER: 1486 bsd.policy = SCHED_OTHER; 1487 break; 1488 case LINUX_SCHED_FIFO: 1489 bsd.policy = SCHED_FIFO; 1490 break; 1491 case LINUX_SCHED_RR: 1492 bsd.policy = SCHED_RR; 1493 break; 1494 default: 1495 return (EINVAL); 1496 } 1497 return (sys_sched_get_priority_min(td, &bsd)); 1498} 1499 1500#define REBOOT_CAD_ON 0x89abcdef 1501#define REBOOT_CAD_OFF 0 1502#define REBOOT_HALT 0xcdef0123 1503#define REBOOT_RESTART 0x01234567 1504#define REBOOT_RESTART2 0xA1B2C3D4 1505#define REBOOT_POWEROFF 0x4321FEDC 1506#define REBOOT_MAGIC1 0xfee1dead 1507#define REBOOT_MAGIC2 0x28121969 1508#define REBOOT_MAGIC2A 0x05121996 1509#define REBOOT_MAGIC2B 0x16041998 1510 1511int 1512linux_reboot(struct thread *td, struct linux_reboot_args *args) 1513{ 1514 struct reboot_args bsd_args; 1515 1516#ifdef DEBUG 1517 if (ldebug(reboot)) 1518 printf(ARGS(reboot, "0x%x"), args->cmd); 1519#endif 1520 1521 if (args->magic1 != REBOOT_MAGIC1) 1522 return (EINVAL); 1523 1524 switch (args->magic2) { 1525 case REBOOT_MAGIC2: 1526 case REBOOT_MAGIC2A: 1527 case REBOOT_MAGIC2B: 1528 break; 1529 default: 1530 return (EINVAL); 1531 } 1532 1533 switch (args->cmd) { 1534 case REBOOT_CAD_ON: 1535 case REBOOT_CAD_OFF: 1536 return (priv_check(td, PRIV_REBOOT)); 1537 case REBOOT_HALT: 1538 bsd_args.opt = RB_HALT; 1539 break; 1540 case REBOOT_RESTART: 1541 case REBOOT_RESTART2: 1542 bsd_args.opt = 0; 1543 break; 1544 case REBOOT_POWEROFF: 1545 bsd_args.opt = RB_POWEROFF; 1546 break; 1547 default: 1548 return (EINVAL); 1549 } 1550 return (sys_reboot(td, &bsd_args)); 1551} 1552 1553 1554/* 1555 * The FreeBSD native getpid(2), getgid(2) and getuid(2) also modify 1556 * td->td_retval[1] when COMPAT_43 is defined. This clobbers registers that 1557 * are assumed to be preserved. The following lightweight syscalls fixes 1558 * this. See also linux_getgid16() and linux_getuid16() in linux_uid16.c 1559 * 1560 * linux_getpid() - MP SAFE 1561 * linux_getgid() - MP SAFE 1562 * linux_getuid() - MP SAFE 1563 */ 1564 1565int 1566linux_getpid(struct thread *td, struct linux_getpid_args *args) 1567{ 1568 1569#ifdef DEBUG 1570 if (ldebug(getpid)) 1571 printf(ARGS(getpid, "")); 1572#endif 1573 td->td_retval[0] = td->td_proc->p_pid; 1574 1575 return (0); 1576} 1577 1578int 1579linux_gettid(struct thread *td, struct linux_gettid_args *args) 1580{ 1581 struct linux_emuldata *em; 1582 1583#ifdef DEBUG 1584 if (ldebug(gettid)) 1585 printf(ARGS(gettid, "")); 1586#endif 1587 1588 em = em_find(td); 1589 KASSERT(em != NULL, ("gettid: emuldata not found.\n")); 1590 1591 td->td_retval[0] = em->em_tid; 1592 1593 return (0); 1594} 1595 1596 1597int 1598linux_getppid(struct thread *td, struct linux_getppid_args *args) 1599{ 1600 1601#ifdef DEBUG 1602 if (ldebug(getppid)) 1603 printf(ARGS(getppid, "")); 1604#endif 1605 1606 PROC_LOCK(td->td_proc); 1607 td->td_retval[0] = td->td_proc->p_pptr->p_pid; 1608 PROC_UNLOCK(td->td_proc); 1609 return (0); 1610} 1611 1612int 1613linux_getgid(struct thread *td, struct linux_getgid_args *args) 1614{ 1615 1616#ifdef DEBUG 1617 if (ldebug(getgid)) 1618 printf(ARGS(getgid, "")); 1619#endif 1620 1621 td->td_retval[0] = td->td_ucred->cr_rgid; 1622 return (0); 1623} 1624 1625int 1626linux_getuid(struct thread *td, struct linux_getuid_args *args) 1627{ 1628 1629#ifdef DEBUG 1630 if (ldebug(getuid)) 1631 printf(ARGS(getuid, "")); 1632#endif 1633 1634 td->td_retval[0] = td->td_ucred->cr_ruid; 1635 return (0); 1636} 1637 1638 1639int 1640linux_getsid(struct thread *td, struct linux_getsid_args *args) 1641{ 1642 struct getsid_args bsd; 1643 1644#ifdef DEBUG 1645 if (ldebug(getsid)) 1646 printf(ARGS(getsid, "%i"), args->pid); 1647#endif 1648 1649 bsd.pid = args->pid; 1650 return (sys_getsid(td, &bsd)); 1651} 1652 1653int 1654linux_nosys(struct thread *td, struct nosys_args *ignore) 1655{ 1656 1657 return (ENOSYS); 1658} 1659 1660int 1661linux_getpriority(struct thread *td, struct linux_getpriority_args *args) 1662{ 1663 struct getpriority_args bsd_args; 1664 int error; 1665 1666#ifdef DEBUG 1667 if (ldebug(getpriority)) 1668 printf(ARGS(getpriority, "%i, %i"), args->which, args->who); 1669#endif 1670 1671 bsd_args.which = args->which; 1672 bsd_args.who = args->who; 1673 error = sys_getpriority(td, &bsd_args); 1674 td->td_retval[0] = 20 - td->td_retval[0]; 1675 return (error); 1676} 1677 1678int 1679linux_sethostname(struct thread *td, struct linux_sethostname_args *args) 1680{ 1681 int name[2]; 1682 1683#ifdef DEBUG 1684 if (ldebug(sethostname)) 1685 printf(ARGS(sethostname, "*, %i"), args->len); 1686#endif 1687 1688 name[0] = CTL_KERN; 1689 name[1] = KERN_HOSTNAME; 1690 return (userland_sysctl(td, name, 2, 0, 0, 0, args->hostname, 1691 args->len, 0, 0)); 1692} 1693 1694int 1695linux_setdomainname(struct thread *td, struct linux_setdomainname_args *args) 1696{ 1697 int name[2]; 1698 1699#ifdef DEBUG 1700 if (ldebug(setdomainname)) 1701 printf(ARGS(setdomainname, "*, %i"), args->len); 1702#endif 1703 1704 name[0] = CTL_KERN; 1705 name[1] = KERN_NISDOMAINNAME; 1706 return (userland_sysctl(td, name, 2, 0, 0, 0, args->name, 1707 args->len, 0, 0)); 1708} 1709 1710int 1711linux_exit_group(struct thread *td, struct linux_exit_group_args *args) 1712{ 1713 1714#ifdef DEBUG 1715 if (ldebug(exit_group)) 1716 printf(ARGS(exit_group, "%i"), args->error_code); 1717#endif 1718 1719 LINUX_CTR2(exit_group, "thread(%d) (%d)", td->td_tid, 1720 args->error_code); 1721 1722 /* 1723 * XXX: we should send a signal to the parent if 1724 * SIGNAL_EXIT_GROUP is set. We ignore that (temporarily?) 1725 * as it doesnt occur often. 1726 */ 1727 exit1(td, W_EXITCODE(args->error_code, 0)); 1728 /* NOTREACHED */ 1729} 1730 1731#define _LINUX_CAPABILITY_VERSION 0x19980330 1732 1733struct l_user_cap_header { 1734 l_int version; 1735 l_int pid; 1736}; 1737 1738struct l_user_cap_data { 1739 l_int effective; 1740 l_int permitted; 1741 l_int inheritable; 1742}; 1743 1744int 1745linux_capget(struct thread *td, struct linux_capget_args *args) 1746{ 1747 struct l_user_cap_header luch; 1748 struct l_user_cap_data lucd; 1749 int error; 1750 1751 if (args->hdrp == NULL) 1752 return (EFAULT); 1753 1754 error = copyin(args->hdrp, &luch, sizeof(luch)); 1755 if (error != 0) 1756 return (error); 1757 1758 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1759 luch.version = _LINUX_CAPABILITY_VERSION; 1760 error = copyout(&luch, args->hdrp, sizeof(luch)); 1761 if (error) 1762 return (error); 1763 return (EINVAL); 1764 } 1765 1766 if (luch.pid) 1767 return (EPERM); 1768 1769 if (args->datap) { 1770 /* 1771 * The current implementation doesn't support setting 1772 * a capability (it's essentially a stub) so indicate 1773 * that no capabilities are currently set or available 1774 * to request. 1775 */ 1776 bzero (&lucd, sizeof(lucd)); 1777 error = copyout(&lucd, args->datap, sizeof(lucd)); 1778 } 1779 1780 return (error); 1781} 1782 1783int 1784linux_capset(struct thread *td, struct linux_capset_args *args) 1785{ 1786 struct l_user_cap_header luch; 1787 struct l_user_cap_data lucd; 1788 int error; 1789 1790 if (args->hdrp == NULL || args->datap == NULL) 1791 return (EFAULT); 1792 1793 error = copyin(args->hdrp, &luch, sizeof(luch)); 1794 if (error != 0) 1795 return (error); 1796 1797 if (luch.version != _LINUX_CAPABILITY_VERSION) { 1798 luch.version = _LINUX_CAPABILITY_VERSION; 1799 error = copyout(&luch, args->hdrp, sizeof(luch)); 1800 if (error) 1801 return (error); 1802 return (EINVAL); 1803 } 1804 1805 if (luch.pid) 1806 return (EPERM); 1807 1808 error = copyin(args->datap, &lucd, sizeof(lucd)); 1809 if (error != 0) 1810 return (error); 1811 1812 /* We currently don't support setting any capabilities. */ 1813 if (lucd.effective || lucd.permitted || lucd.inheritable) { 1814 linux_msg(td, 1815 "capset effective=0x%x, permitted=0x%x, " 1816 "inheritable=0x%x is not implemented", 1817 (int)lucd.effective, (int)lucd.permitted, 1818 (int)lucd.inheritable); 1819 return (EPERM); 1820 } 1821 1822 return (0); 1823} 1824 1825int 1826linux_prctl(struct thread *td, struct linux_prctl_args *args) 1827{ 1828 int error = 0, max_size; 1829 struct proc *p = td->td_proc; 1830 char comm[LINUX_MAX_COMM_LEN]; 1831 struct linux_emuldata *em; 1832 int pdeath_signal; 1833 1834#ifdef DEBUG 1835 if (ldebug(prctl)) 1836 printf(ARGS(prctl, "%d, %d, %d, %d, %d"), args->option, 1837 args->arg2, args->arg3, args->arg4, args->arg5); 1838#endif 1839 1840 switch (args->option) { 1841 case LINUX_PR_SET_PDEATHSIG: 1842 if (!LINUX_SIG_VALID(args->arg2)) 1843 return (EINVAL); 1844 em = em_find(td); 1845 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1846 em->pdeath_signal = args->arg2; 1847 break; 1848 case LINUX_PR_GET_PDEATHSIG: 1849 em = em_find(td); 1850 KASSERT(em != NULL, ("prctl: emuldata not found.\n")); 1851 pdeath_signal = em->pdeath_signal; 1852 error = copyout(&pdeath_signal, 1853 (void *)(register_t)args->arg2, 1854 sizeof(pdeath_signal)); 1855 break; 1856 case LINUX_PR_GET_KEEPCAPS: 1857 /* 1858 * Indicate that we always clear the effective and 1859 * permitted capability sets when the user id becomes 1860 * non-zero (actually the capability sets are simply 1861 * always zero in the current implementation). 1862 */ 1863 td->td_retval[0] = 0; 1864 break; 1865 case LINUX_PR_SET_KEEPCAPS: 1866 /* 1867 * Ignore requests to keep the effective and permitted 1868 * capability sets when the user id becomes non-zero. 1869 */ 1870 break; 1871 case LINUX_PR_SET_NAME: 1872 /* 1873 * To be on the safe side we need to make sure to not 1874 * overflow the size a linux program expects. We already 1875 * do this here in the copyin, so that we don't need to 1876 * check on copyout. 1877 */ 1878 max_size = MIN(sizeof(comm), sizeof(p->p_comm)); 1879 error = copyinstr((void *)(register_t)args->arg2, comm, 1880 max_size, NULL); 1881 1882 /* Linux silently truncates the name if it is too long. */ 1883 if (error == ENAMETOOLONG) { 1884 /* 1885 * XXX: copyinstr() isn't documented to populate the 1886 * array completely, so do a copyin() to be on the 1887 * safe side. This should be changed in case 1888 * copyinstr() is changed to guarantee this. 1889 */ 1890 error = copyin((void *)(register_t)args->arg2, comm, 1891 max_size - 1); 1892 comm[max_size - 1] = '\0'; 1893 } 1894 if (error) 1895 return (error); 1896 1897 PROC_LOCK(p); 1898 strlcpy(p->p_comm, comm, sizeof(p->p_comm)); 1899 PROC_UNLOCK(p); 1900 break; 1901 case LINUX_PR_GET_NAME: 1902 PROC_LOCK(p); 1903 strlcpy(comm, p->p_comm, sizeof(comm)); 1904 PROC_UNLOCK(p); 1905 error = copyout(comm, (void *)(register_t)args->arg2, 1906 strlen(comm) + 1); 1907 break; 1908 default: 1909 error = EINVAL; 1910 break; 1911 } 1912 1913 return (error); 1914} 1915 1916int 1917linux_sched_setparam(struct thread *td, 1918 struct linux_sched_setparam_args *uap) 1919{ 1920 struct sched_param sched_param; 1921 struct thread *tdt; 1922 int error; 1923 1924#ifdef DEBUG 1925 if (ldebug(sched_setparam)) 1926 printf(ARGS(sched_setparam, "%d, *"), uap->pid); 1927#endif 1928 1929 error = copyin(uap->param, &sched_param, sizeof(sched_param)); 1930 if (error) 1931 return (error); 1932 1933 tdt = linux_tdfind(td, uap->pid, -1); 1934 if (tdt == NULL) 1935 return (ESRCH); 1936 1937 error = kern_sched_setparam(td, tdt, &sched_param); 1938 PROC_UNLOCK(tdt->td_proc); 1939 return (error); 1940} 1941 1942int 1943linux_sched_getparam(struct thread *td, 1944 struct linux_sched_getparam_args *uap) 1945{ 1946 struct sched_param sched_param; 1947 struct thread *tdt; 1948 int error; 1949 1950#ifdef DEBUG 1951 if (ldebug(sched_getparam)) 1952 printf(ARGS(sched_getparam, "%d, *"), uap->pid); 1953#endif 1954 1955 tdt = linux_tdfind(td, uap->pid, -1); 1956 if (tdt == NULL) 1957 return (ESRCH); 1958 1959 error = kern_sched_getparam(td, tdt, &sched_param); 1960 PROC_UNLOCK(tdt->td_proc); 1961 if (error == 0) 1962 error = copyout(&sched_param, uap->param, 1963 sizeof(sched_param)); 1964 return (error); 1965} 1966 1967/* 1968 * Get affinity of a process. 1969 */ 1970int 1971linux_sched_getaffinity(struct thread *td, 1972 struct linux_sched_getaffinity_args *args) 1973{ 1974 int error; 1975 struct thread *tdt; 1976 struct cpuset_getaffinity_args cga; 1977 1978#ifdef DEBUG 1979 if (ldebug(sched_getaffinity)) 1980 printf(ARGS(sched_getaffinity, "%d, %d, *"), args->pid, 1981 args->len); 1982#endif 1983 if (args->len < sizeof(cpuset_t)) 1984 return (EINVAL); 1985 1986 tdt = linux_tdfind(td, args->pid, -1); 1987 if (tdt == NULL) 1988 return (ESRCH); 1989 1990 PROC_UNLOCK(tdt->td_proc); 1991 cga.level = CPU_LEVEL_WHICH; 1992 cga.which = CPU_WHICH_TID; 1993 cga.id = tdt->td_tid; 1994 cga.cpusetsize = sizeof(cpuset_t); 1995 cga.mask = (cpuset_t *) args->user_mask_ptr; 1996 1997 if ((error = sys_cpuset_getaffinity(td, &cga)) == 0) 1998 td->td_retval[0] = sizeof(cpuset_t); 1999 2000 return (error); 2001} 2002 2003/* 2004 * Set affinity of a process. 2005 */ 2006int 2007linux_sched_setaffinity(struct thread *td, 2008 struct linux_sched_setaffinity_args *args) 2009{ 2010 struct cpuset_setaffinity_args csa; 2011 struct thread *tdt; 2012 2013#ifdef DEBUG 2014 if (ldebug(sched_setaffinity)) 2015 printf(ARGS(sched_setaffinity, "%d, %d, *"), args->pid, 2016 args->len); 2017#endif 2018 if (args->len < sizeof(cpuset_t)) 2019 return (EINVAL); 2020 2021 tdt = linux_tdfind(td, args->pid, -1); 2022 if (tdt == NULL) 2023 return (ESRCH); 2024 2025 PROC_UNLOCK(tdt->td_proc); 2026 csa.level = CPU_LEVEL_WHICH; 2027 csa.which = CPU_WHICH_TID; 2028 csa.id = tdt->td_tid; 2029 csa.cpusetsize = sizeof(cpuset_t); 2030 csa.mask = (cpuset_t *) args->user_mask_ptr; 2031 2032 return (sys_cpuset_setaffinity(td, &csa)); 2033} 2034 2035int 2036linux_sched_rr_get_interval(struct thread *td, 2037 struct linux_sched_rr_get_interval_args *uap) 2038{ 2039 struct timespec ts; 2040 struct l_timespec lts; 2041 struct thread *tdt; 2042 int error; 2043 2044 tdt = linux_tdfind(td, uap->pid, -1); 2045 if (tdt == NULL) 2046 return (ESRCH); 2047 2048 error = kern_sched_rr_get_interval_td(td, tdt, &ts); 2049 PROC_UNLOCK(tdt->td_proc); 2050 if (error != 0) 2051 return (error); 2052 lts.tv_sec = ts.tv_sec; 2053 lts.tv_nsec = ts.tv_nsec; 2054 return (copyout(<s, uap->interval, sizeof(lts))); 2055} 2056 2057/* 2058 * In case when the Linux thread is the initial thread in 2059 * the thread group thread id is equal to the process id. 2060 * Glibc depends on this magic (assert in pthread_getattr_np.c). 2061 */ 2062struct thread * 2063linux_tdfind(struct thread *td, lwpid_t tid, pid_t pid) 2064{ 2065 struct linux_emuldata *em; 2066 struct thread *tdt; 2067 struct proc *p; 2068 2069 tdt = NULL; 2070 if (tid == 0 || tid == td->td_tid) { 2071 tdt = td; 2072 PROC_LOCK(tdt->td_proc); 2073 } else if (tid > PID_MAX) 2074 tdt = tdfind(tid, pid); 2075 else { 2076 /* 2077 * Initial thread where the tid equal to the pid. 2078 */ 2079 p = pfind(tid); 2080 if (p != NULL) { 2081 if (SV_PROC_ABI(p) != SV_ABI_LINUX) { 2082 /* 2083 * p is not a Linuxulator process. 2084 */ 2085 PROC_UNLOCK(p); 2086 return (NULL); 2087 } 2088 FOREACH_THREAD_IN_PROC(p, tdt) { 2089 em = em_find(tdt); 2090 if (tid == em->em_tid) 2091 return (tdt); 2092 } 2093 PROC_UNLOCK(p); 2094 } 2095 return (NULL); 2096 } 2097 2098 return (tdt); 2099} 2100 2101void 2102linux_to_bsd_waitopts(int options, int *bsdopts) 2103{ 2104 2105 if (options & LINUX_WNOHANG) 2106 *bsdopts |= WNOHANG; 2107 if (options & LINUX_WUNTRACED) 2108 *bsdopts |= WUNTRACED; 2109 if (options & LINUX_WEXITED) 2110 *bsdopts |= WEXITED; 2111 if (options & LINUX_WCONTINUED) 2112 *bsdopts |= WCONTINUED; 2113 if (options & LINUX_WNOWAIT) 2114 *bsdopts |= WNOWAIT; 2115 2116 if (options & __WCLONE) 2117 *bsdopts |= WLINUXCLONE; 2118} 2119